Does metal have a GPU timeout?

[FilippoLeon]

So I have a weird regression on some very complex fragment shader on intel GPU macs after the SOnoma update.

Basically I get the following crash:
Execution of the command buffer was aborted due to an error during execution. Caused GPU Timeout Error (00000002:kIOAccelCommandBufferCallbackErrorTimeout)
[mvk-error] VK_ERROR_DEVICE_LOST: MTLCommandBuffer "vkQueueSubmit CommandBuffer on Queue 0-0" execution failed (code 2): Caused GPU Timeout Error (00000002:kIOAccelCommandBufferCallbackErrorTimeout)

and the validation is reporting no errors.

It seems that the shader is somehow too complex for the GPU to handle and Metal is killing my shader?

Any idea how to further debug this? Are there any ways to increase the timeout time if the problem is that? (I know windows has TDR delays)

Thanks a loh.